On the Specification of Propensity Scores: with an Application to the WTO-Environment Debate

The use of propensity score methods for program evaluation with non-experimental data typically requires the propensity score be estimated, often with a model whose specification is unknown. While theoretical results suggest that estimators utilizing more flexible propensity score specifications perform better, this has not filtered into applied research. Here, we provide Monte Carlo evidence indicating the benefits of over-specifying the propensity score when using weighting estimators, as well as using normalized weights. We illustrate these results with an application assessing the environmental effects of GATT/WTO membership. We find that membership has a mixed impact, and that under-fitting the propensity score yields misleading inference in several cases.